The sequence section starts with a number, which tells how sequences are present in the section.
If this number if 0, the section automatically ends.
The number 0 can be represented using the 1 byte or the 2 bytes formats.
That's because the 2-bytes formats fully overlaps the 1 byte format.
However, when 0 is represented using the 2-bytes format,
the decoder was expecting the sequence section to continue,
and was looking for FSE tables, which is incorrect.
Fixed this behavior, in both the reference decoder and the educational behavior.
In practice, this behavior never happens,
because the encoder will always select the 1-byte format to represent 0,
since this is more efficient.
Completed the fix with a new golden sample for tests,
a clarification of the specification,
and a decoder errata paragraph.
and in `decodecorpus`:
the specific case `nbSeq=127` can be represented using the 1-byte format.
Note that both the 1-byte and the 2-bytes formats are valid to represent this case,
so there was no "error", produced data remains valid,
it's just that the 1-byte format is more efficient.
fix#3667
Credit to @ip7z for finding this issue.
Reduces memory when blocks are guaranteed to be smaller than allowed by
the format. This is useful for streaming compression in conjunction with
ZSTD_c_maxBlockSize.
This PR saves 2 * (formatMaxBlockSize - paramMaxBlockSize) when streaming.
Once it is rebased on top of PR #3616 it will save
3 * (formatMaxBlockSize - paramMaxBlockSize).
The split literals buffer patch increased streaming decompression memory
by 64KB (shrunk lit buffer from 128KB to 64KB, and added 128KB). This
patch removes the added 128KB buffer, because it isn't necessary.
The buffer was there because the literals compression code didn't know
the true `blockSizeMax` of the frame, and always put split literals so
they ended 128KB - 32 from the beginning of the block. Instead, we can
pass down the true `blockSizeMax` and ensure that the split literals
end up at `blockSizeMax - 32` from the beginning of the block. We
already reserve a full `blockSizeMax` bytes in streaming mode, so we
won't be overwriting the extDict window.
detected by @terrelln,
these issue could be triggered in specific scenarios
namely decompression of certain invalid magic-less frames,
or requested properties from certain invalid skippable frames.
* add check for valid dest buffer and fuzz on random dest ptr when malloc 0
* add uptrval to linux-kernel
* remove bin files
* get rid of uptrval
* restrict max pointer value check to platforms where sizeof(size_t) == sizeof(void*)
Looking at the __builtin_expect in ZSTD_decodeSequence:
{ size_t offset;
#if defined(__clang__)
if (LIKELY(ofBits > 1)) {
#else
if (ofBits > 1) {
#endif
ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
From profile-annotated assembly, the probability of ofBits > 1 is about 75%
(101k counts out of 135k counts). This is much smaller than the recommended
likelihood to use __builtin_expect which is 99%. As a result, clang moved the
else block further away which hurts cache locality. Removing this
__built_expect along with two others in ZSTD_decodeSequence gave better
performance when PGO is enabled. I suggest to remove these branch hints and
rely on PGO which leverages runtime profiles from actual workload to calculate
branch probability instead.
Inlining `BIT_reloadDStream` provided >3% decompression speed improvement for
clang PGO-optimized zstd binary, measured using the Silesia corpus with
compression level 1. The win comes from improved register allocation which leads
to fewer spills and reloads. Take a look at this comparison of
profile-annotated hot assembly before and after this change:
https://www.diffchecker.com/UjDGIyLz/. The diff is a bit messy, but notice three
fewer moves after inlining.
In general LLVM's register allocator works better when it can see more code. For
example, when the register allocator sees a call instruction, it partitions the
registers into caller registers and callee registers, and it is not free to do
whatever it wants with all the registers for the current function. Inlining the
callee lets the register allocation access all registers and use them more
flexsibly.
Every 256 bytes the lazy match finders process without finding a match,
they will increase their step size by 1. So for bytes [0, 256) they search
every position, for bytes [256, 512) they search every other position,
and so on. However, they currently still insert every position into
their hash tables. This is different from fast & dfast, which only
insert the positions they search.
This PR changes that, so now after we've searched 2KB without finding
any matches, at which point we'll only be searching one in 9 positions,
we'll stop inserting every position, and only insert the positions we
search. The exact cutoff of 2KB isn't terribly important, I've just
selected a cutoff that is reasonably large, to minimize the impact on
"normal" data.
This PR only adds skipping to greedy, lazy, and lazy2, but does not
touch btlazy2.
| Dataset | Level | Compiler | CSize ∆ | Speed ∆ |
|---------|-------|--------------|---------|---------|
| Random | 5 | clang-14.0.6 | 0.0% | +704% |
| Random | 5 | gcc-12.2.0 | 0.0% | +670% |
| Random | 7 | clang-14.0.6 | 0.0% | +679% |
| Random | 7 | gcc-12.2.0 | 0.0% | +657% |
| Random | 12 | clang-14.0.6 | 0.0% | +1355% |
| Random | 12 | gcc-12.2.0 | 0.0% | +1331% |
| Silesia | 5 | clang-14.0.6 | +0.002% | +0.35% |
| Silesia | 5 | gcc-12.2.0 | +0.002% | +2.45% |
| Silesia | 7 | clang-14.0.6 | +0.001% | -1.40% |
| Silesia | 7 | gcc-12.2.0 | +0.007% | +0.13% |
| Silesia | 12 | clang-14.0.6 | +0.011% | +22.70% |
| Silesia | 12 | gcc-12.2.0 | +0.011% | -6.68% |
| Enwik8 | 5 | clang-14.0.6 | 0.0% | -1.02% |
| Enwik8 | 5 | gcc-12.2.0 | 0.0% | +0.34% |
| Enwik8 | 7 | clang-14.0.6 | 0.0% | -1.22% |
| Enwik8 | 7 | gcc-12.2.0 | 0.0% | -0.72% |
| Enwik8 | 12 | clang-14.0.6 | 0.0% | +26.19% |
| Enwik8 | 12 | gcc-12.2.0 | 0.0% | -5.70% |
The speed difference for clang at level 12 is real, but is probably
caused by some sort of alignment or codegen issues. clang is
significantly slower than gcc before this PR, but gets up to parity with
it.
I also measured the ratio difference for the HC match finder, and it
looks basically the same as the row-based match finder. The speedup on
random data looks similar. And performance is about neutral, without the
big difference at level 12 for either clang or gcc.
* Mark all bufferless and block level functions as deprecated
* Update documentation to suggest not using these functions
* Add `_deprecated()` wrappers for functions that we use internally and
call those instead
